12. Biostatistics Flashcards
Define “parametric tests” and give examples
Interval or ratio data; tests hypothesis in NORMALLY distributed population. MORE POWER.
Examples of parametric tests
t-test; ANOVA; Pearson Coefficient; Multiple Regression
Define “non-parametric tests” and give examples
Data from a SKEWED distribution; LESS POWER
Examples of non-parametric tests
Chi-Square; Logistic regression; Spearman Rho; Wilcoxon Signed Rank
Define “degrees of freedom”
of pieces of info that can vary independently from one another; n -1, where n is # of cases in the sample
One-tailed t-test
Used when direction of difference is known or postulated
Two-tailed t-test
Used when direction of difference is unknown or not postulated
Paired t-test
BEFORE & AFTER tests of SAME GROUP; Pre & Post test analysis
Independent t-test
Most common statistical method; Simple (yet powerful); Comparison between control and test group. ONLY TWO GROUPS CAN BE COMPARED. Assumes normal distribution. AKA = two-sample t-test; tests DIFFERENT GROUPS
Analysis of Variance (ANOVA)
Accepted method of comparing TWO OR MORE groups from ONE STUDY; determines RELATIONSHIP BETWEEN IV & DV; Test statistic is an F RATIO. Variables are factors.
Post-hoc tests
Defines which of 3 or more groups are actually different; test after ANOVA to detemine what the significance actually is
Tukey
Post-hoc test; Used when larger number of comparisons are made; assumes groups are of equal size; Used to determine which means are significantly different
Dunn
Post-hoc test; Used when only a few comparisons are made
Bonferroni
Post-hoc test; Used when 5 or less comparisons are made
Repeated Measures ANOVA
Used when experiment involves matched subjects; Measure an outcome in each subject before, during, and after intervention
Analysis of Covariance (ANCOVA)
Enables testing of differences among at least 3 groups, while adjusting for effects of covariates or confounders on DV
MANOVA
Used when 2 DVs are assessed
Chi-Square Test
Non-parametric; applied to non-normally distributed populations; “Goodness of Fit test”or “Pearson Chi Square”; Statistically significant difference between observed (actual) frequencies and expected frequency of variables; Answers research questions about RATES, PROPORTIONS, OR FREQUENCIES. GREATER CHI SQUARE, LESS LIKELY DIFFERENCE IS D/T CHANCE
Fisher’s Exact Test
Non-parametric test for nominal data. Used instead of Chi Square in VERY SMALL SAMPLE SIZE
McNemar’s Test
Non-parametric test for nominal data from PAIRED SAMPLES
Mantel-Haenszel
Non-parametric test for nominal data to control for effects of a confounder.
Mann-Whitney U Test
Non-parametric test for ordinal variables; OFTEN USED TO COMPARE SURVIVAL CURVES
Wilcoxon Rank Signed Test
Non-parametric equivalent of PAIRED T-TEST for ORDINAL variables. Use with 2 correlated samples (before & after) or difference between 2 groups.
Kruskal-Wallis test
Non-parametric equivalent of ANOVA to compare ? 3 groups of different groups with ordinal data
Friedman’s test
Non-parametric equivalent of REPEATED MEASURES ANOVA to compare ? 3 groups of RELATED groups with ordinal data
Correlation
Quantitative way of measuring the strength of a relationship between two variables; DOES NOT ASSUME CAUSE & EFFECT. 0 = NO RELATIONSHIP. +1 = DEFINITE POSITIVE CORRELATION; -1 = DEFINITE INVERSE RELATIONSHIP
Pearson’s Correlation Coefficient
Parametric measure of association
Spearman’s Rho Coefficient
Nonparametric measure of correlation between 2 quantitative ORDINAL variables
Regression
Prediction of one variable from another; Assumes cause & effect relationship. Y = mX + b
Multiple Regression
Same as simple regression, but INCLUDES MULTIPLE INDEPENDENT VARIABLES and 1 DEPENDENT VARIABLE. Predictor values (IV) to predict as single DV (criterion variable)
Logistic Regression
Similar to multiple regression. DV is categorical (ordinal). SHOULD NOT INCLUDE CONTINUOUS VARIABLES.
Cronbach Alpha
Index of internal consistency, reliability. Degree to which responses are consistent across multiple measures of same construct.
Homogeneity tests
Assumption variances of population being compared with t-test or ANOVA are equal. Assumes equal populations. Possible violations detected with LEVINE’S TEST.
Levine’s test
Used prior to conducting ANOVA and before interpretting results of t-tests
Limitations
Influences researcher CANNOT CONTROL. Potential weakness of study.
Delimitations
Influences that could be controlled, but weren’t.
Bias
Preference toward a result. May be d/t SAMPLE selection; who is reading the results; BLINDED STUDIES
